How to Test Your Brand in AI: Branded, Category and Problem-First Questions

How do you test your brand in AI?

Test your brand in AI by asking a balanced set of questions about your company, its category and the customer problems it addresses.

Compare the responses with your substantiated positioning. Examine both how the company is described and whether it appears in relevant comparisons and recommendations.

Repeat the questions under comparable conditions and record the results. One response can reveal an issue to investigate; it cannot establish a reliable pattern.

Start with the customer's question

A common starting point is to ask an AI assistant:

"What do you know about our company?"

That is useful. It can reveal factual errors, outdated descriptions and missing capabilities.

But the question already supplies your name. It does not show whether the assistant includes your company when a buyer asks about a need without naming a provider.

A useful assessment therefore includes both branded and unbranded questions.

Begin by identifying plausible buying situations:

• What problem is the customer trying to solve?
• What kind of provider could help?
• Which requirements affect the choice?
• In which market or geography must the provider operate?
• What would make your company a credible option?

Use customer interviews, sales conversations, enquiries and available search data to inform these situations. Internally invented questions are a starting hypothesis, not evidence of actual buyer behaviour.

Establish what an accurate answer should contain

Before testing, document Intended Brand Understanding.

Record the company's category, audience, capabilities, differentiation and supporting evidence. Include the limits of its offering.

This reference helps distinguish an inaccurate answer from one that simply uses different wording.

It also prevents the assessment from becoming a search for flattering language.

If your company wants to be described as a specialist, specify the capabilities and evidence that justify that description. If it serves a particular market, make its availability and fit clear.

Five question types to include

1. Branded questions

Branded questions name the company.

Examples:

"What does [Company] offer, and which customers does it serve?"

"What are [Company]'s main capabilities and limitations?"

"What evidence supports [Company]'s claimed expertise in [area]?"

Use these to examine identity, category, capabilities, audience and substantiation.

Avoid assuming the desired conclusion. "Why is [Company] the best provider?" encourages a favourable response rather than testing whether that position is supported.

2. Category questions

Category questions examine how the assistant describes the market and its participants.

Examples:

"What types of providers operate in [category]?"

"Which companies offer [capability] in [market]?"

"How do providers in [category] differ?"

Check whether the category matches how customers actually frame the decision. An internal category label may not be the language buyers use.

3. Problem-first questions

Problem-first questions begin with a customer need.

Examples:

"How can a [type of organisation] address [specific problem]?"

"What approaches could reduce [operational difficulty] under [constraints]?"

"Which providers could help with [problem] in [geography]?"

These questions can reveal whether the company is associated with a relevant need, and whether the assistant recommends a different type of solution altogether.

Do not assume that every problem should lead to a supplier recommendation. Advice, an internal process change or another approach may be appropriate.

4. Comparison questions

Comparison questions examine competitive representation.

Examples:

"How do [Company A] and [Company B] differ for [use case]?"

"What trade-offs should a buyer consider between these providers?"

"Which requirements would make one option more suitable than another?"

Use genuinely relevant alternatives. Record whether the answer makes unsupported claims, overlooks important differences or compares offerings that are not equivalent.

5. Recommendation questions

Recommendation questions describe a buying situation and request suitable options.

Examples:

"Which providers should a [customer type] consider for [need], given [requirements]?"

"Suggest a shortlist for [use case] in [market], and explain the basis for inclusion."

"What information would you need before recommending a provider for this requirement?"

Examine fit as well as inclusion. Being recommended for work the company cannot deliver is not a positive result.

An illustrative prompt set

Consider a hypothetical company providing rapid microbial testing technology and implementation support.

A balanced initial set could include:

Branded:

"What does [Company] offer, and which laboratories does it serve?"

Category:

"Which providers offer rapid microbial testing technology for pharmaceutical quality-control laboratories in [market]?"

Problem-first:

"What approaches could a pharmaceutical laboratory consider to reduce microbial testing time while meeting its validation requirements?"

Comparison:

"How do [Company] and [Relevant Alternative] differ in technology, implementation support and suitability for [application]?"

Recommendation:

"Which providers should a pharmaceutical quality-control laboratory evaluate for [specific application], given [technical and geographic requirements]?"

These are research examples, not technical or regulatory guidance. The scenario is hypothetical and does not describe a KVD client engagement.

Design the test before collecting answers

Choose platforms relevant to your customers and document the scope.

For an initial exploratory assessment, a team could select three plausible buying situations, write questions across the five types, and repeat them in fresh sessions.

This is a practical starting design, not a statistically validated sample size.

Keep an exact-wording test set for comparisons over time. Use a separate set of wording variations to explore how sensitive the results are to phrasing.

Do not silently change the questions between assessments and then attribute all differences to changes in the brand's representation.

Record enough context to interpret the result

Cited sources provide useful leads. They do not necessarily reveal every influence on the answer.

Review understanding and consideration separately

For understanding, ask:

• Is the company identified correctly?
• Are its category and capabilities accurate?
• Does the answer recognise relevant audiences and needs?
• Is differentiation supported by evidence?
• Are important limitations represented?
• Are comparisons appropriate?

For consideration, ask:

• Did the company appear in an unbranded buying situation?
• Was it a credible option under the stated requirements?
• Which alternatives appeared?
• What reasons were given for inclusion or exclusion?
• Did the pattern recur across tests?

These questions correspond to KVD's distinction between the Understanding Gap and the Consideration Gap.

Summarise observations without overstating them

Use counts with clear denominators.

For example:

"The company appeared in 8 of 20 sampled recommendation responses for the defined buying situations."

This describes the test sample. It does not mean the company has 40% of the market's AI visibility or appears in 40% of actual customer conversations.

Likewise, report specific representation findings:

"Six responses described the discontinued service as a current offering."

Preserve differences between platforms and buying situations. A single combined figure can hide where the problem occurs.

If you use a scoring rubric, explain its criteria and judgement calls. Do not present an internal score as an established industry measure.

Turn recurring findings into priorities

A useful findings summary connects each observation to evidence and a next step.

These are starting points for investigation. They do not establish which source caused a response or guarantee that an intervention will change it.

Repeat the assessment and examine changes

After making changes, repeat the comparable test set.

Record what changed in the company's information and what changed in the testing environment. Platforms, models, retrieval behaviour and competing information may also have evolved.

Look for improvements in factual accuracy, relevant differentiation and appropriate inclusion.

Where possible, connect the findings with customer research, qualified enquiries and commercial outcomes. Better sampled answers alone do not establish increased sales or stronger brand equity.

The value of a balanced test

Branded questions show how the assistant describes your company when its name is supplied.

Category, problem-first and recommendation questions help examine whether it appears in relevant buying situations. Comparison questions add context about alternatives and trade-offs.

Together, they provide a more useful basis for action than a collection of isolated screenshots.

Start with a defined brand reference, ask plausible customer questions, and preserve enough evidence to review the results.